Where it all starts

13 reasons why… but make it 2

Not knowing how to initiate my report, (as it happens most of the time with school essays), I thought that telling you why I chose the topic I did could be a good starting point. Well, before explaining to you why maybe I should start with what. In this report, you will follow the path I took aiming to see if football (or soccer depending on where you are from) teams win more games when playing at home or away. As a big football fan, I am pleased to be able to work on a topic that is interesting and fascinating to me. I guess that’s the main reason I picked this topic. Searching for football data, I discovered there were lots of choices, there is data on players, games, countries, goals scored, and so on. As I often go to see Servette football games when they are playing at home but also away, I wonder if the probabilities of seeing them win would be different in each case. I guess those are the two main reasons I decided to go in the direction I did, and found my research question. Indeed, I wonder if football teams win more at home or away.

Servette Football Fans during a game at Home (Stade de la Praille)

Adventuring into Data Collection

As I mentioned below, there is a lot of data revolving around the football world. When I stumbled across the website Football Data I was at first a bit reluctant as it looks more like a betting page than anything else and is fill with advertising of different betting houses.

Screen Shot from the Football Data Website

I don’t think I was totally wrong, but I must admit their data sets on football are pretty complete, with lots of information and have data from a lot of countries. After thinking a bit more about what I wanted to do, I finally decided to focus my research on Switzerland, to be more precise on the Super League championship. I found a complete data set that listed all the football games played in the Swiss Super League from 2012 up to now. I was glad to find that the data was updated every week. Before dowloading it, I decided to wait until the 13th of November 2022, which is when the winter football break started, and meant no more games until 2023. By doing so I now have an up-to-date data set and I am ready to start exploring it (plus to be fair I wouldn’t have started before, I am more of a kind of last-minute and no-sleep-nights person).

It actually isn’t anymore but the cat was cute

DataBase_OnFootballScores <- read_excel("DataBase_OnFootballScores.xlsx")
DBF<-DataBase_OnFootballScores

Exploring my data

I started by doing a table of both the games played at Home and those played Away with the function table so I could visualize a bit more my data.I also added a “word cloud” because I thougt it was fun.

## 
##        Aarau        Basel Grasshoppers     Lausanne       Lugano       Luzern 
##           37          188          152          108          134          188 
## Schaffhausen     Servette         Sion   St. Gallen         Thun        Vaduz 
##            1           80          189          188          146           73 
##   Winterthur        Xamax   Young Boys       Zurich 
##            8           37          188          170

## 
##        Aarau        Basel Grasshoppers     Lausanne       Lugano       Luzern 
##           37          188          152          108          134          189 
##     Servette         Sion   St. Gallen         Thun        Vaduz   Winterthur 
##           80          189          188          146           73            8 
##        Xamax   Young Boys       Zurich 
##           37          188          170

I noticed that Schaffhausen had only one game which seemed strange to me. I search for the game on Google using the date and found out that it was a playoff game. I decided to take it off the analysis as it isn’t really a game in the Super League championship. Two other games were deleted as they appear to be from the Challenge League and not the Super League. I then made sure that all remaining games were played in Super League during the championship. To delete the data I didn’t need, I created a new data set called DBF2 using the filter function in the dplyr package.

DBF2<- DBF %>% filter(DBF$League== "Super League", DBF$Away!="Schaffhausen") 

DBF2

(Not so) Fun Fact: I spend like 15 minutes desperate because the code didn’t work only to notice that the problem was that I hadn’t written “Schaffhausen” correctly.

## [1] "😢"

A difficult path

Finding the Hypothesis

Once my data was all set (lol), I started working properly on it. First I set up my null and alternative hypotheses. I changed my hypothesis a lot during the process until I could find one that fitted my data and ideas. It wasn’t an easy task, and after spending too much time trying to find the good way to use the statistic I wanted to (and miserably failing) I finally (a day before the due date lol) found the hypothesis that was perfect for my goal. What do I mean by perfect ? Well that fitted my ideas and was doable with my data and knowledge. My null hypothesis is that, the number of wins at Home is the same as the number of wins Away in the Swiss Super League championship. My alternative hypothesis is that the number of wins at Home is bigger than the number of wins away in the Swiss Super League championship. Now that my hypotheses are also set, let’s start with the statistics. In statistics you can never say that a hypothesis is right but you can say that it’s wrong. That’s why I will use a chi-square test with the aim to reject my null hypothesis. Basically the chi-square test consists in seeing if the differences observed between two results are statistically different from what would have been expected by chance. My alternative hypothesis is oriented towards more wins at home as it is commonly supposed that teams usually win more oftener at home than away. I actually discover there was a name for this phenomenon, it’s called Home Advantage. This might be because players know better the field, they are not tired from an away trip and might feel more confident. Moreover, there are usually more fans when games are at home than away which could increase the motivation of the players to show off and win the game. In this work, I want to see if those common beliefs are true in the Swiss Super League too or are just old rumors.

ggplot(DBF2 %>% filter(Res != "D")) +
  aes(x = Home, fill = Res) +
  geom_bar() +
  scale_fill_manual(values=c("#FF0066",
                             "#0099CC"))+
  theme(axis.text.x = element_text(angle = 90, vjust = 0.5, hjust=1))

In this graph we can see in blue (H) the number of times the team won the game when playing at home. In pink (A) it’s the number of times it lost when playing at home, so in other words the number of time the Away team won.

ggplot(DBF2 %>% filter(Res != "D")) +
  aes(x = Away, fill = Res) +
  geom_bar() +
  scale_fill_manual(values=c("#FF9966",
                             "#66CC66"))+
  theme(axis.text.x = element_text(angle = 90, vjust = 0.5, hjust=1))

Here we have a barplot with the name of the teams that played away. In orange (A) we see the times the away team won, so the one on the x axis. In green it’s the number of time the Away team lost, so the Home one won.

What those two figures show us is that in both cases, the team playing at home (blue and green) seems to win oftener than the one playing away. This is just a guessing from the visualization of the data, but to be sure a statistic test is needed.

Managing all this data

As many of the columns in my data set weren’t useful for what I wanted to do, I started by simplifying my data by keeping only the ones I was interested in.

# Drop all the non-relevant columns

DBF3<- DBF2 %>% select(Season, Home, Away, Res)

# Res column 
table(DBF3$Res)
## 
##   A   D   H 
## 593 478 813

I kept the columns Home, Away and Res. The columns Home and Away tells which team is at home and which is away, and the column Res gives the outcome of the game. In this last column, when the winning team is the one playing at home it’s coded H, when it’s the one playing away it’s coded A and D is for ties. In order to do a chi-square test, I had to create a new data set.

#Creating a new data set

DBFHR00 <- DBF3 %>% select(Home, Res)
table(DBFHR00)
##               Res
## Home             A   D   H
##   Aarau         19   6  12
##   Basel         20  51 117
##   Grasshoppers  56  33  63
##   Lausanne      59  21  28
##   Lugano        46  35  53
##   Luzern        61  53  74
##   Servette      24  27  29
##   Sion          68  50  70
##   St. Gallen    60  43  85
##   Thun          44  42  59
##   Vaduz         33  26  14
##   Winterthur     3   2   3
##   Xamax         20  10   7
##   Young Boys    23  31 134
##   Zurich        57  48  65
DBFHR0 <- as.data.frame.matrix(table(DBFHR00))

I still have to manage some things in my data before being able to use it for my statistic test, but let’s have a “reading break” and look at some graphs.

ggballoonplot(DBFHR0, col="green",xlab= "Results", ylab="Team playing at Home", main="Result for the games played at Home", width="50%")

This graph show us that some teams have played much less games than others. For example Wintherthur seems to be the team with the less games played. After doing a quick research, I found out that this team has only recently (last season) made it up to the Super League, which explains the few games played. This graph also allow us to see that some teams have in general much more wins than others. I think it’s important to take it into account, because it might bias the results.

Nothing to add

Now that this is done, we are almost to the chi-square test part, we just need some little fixings. In my hypotheses I only mention the wins and losses and leave apart the ties. Surprisingly enough, this isn’t another obstacle in my path to the results, as there actually is a function that lets do this.

DBFHR0
DBFHR <- DBFHR0 %>% select(H, A)
DBFHR

Now that I have all I need in my possession, it’s time to do the chi-square test. The moment of truth as come …

chi2H <- chisq.test(DBFHR)
## Warning in chisq.test(DBFHR): Chi-squared approximation may be incorrect
chi2H
## 
##  Pearson's Chi-squared test
## 
## data:  DBFHR
## X-squared = 152.72, df = 14, p-value < 2.2e-16
sum(DBFHR$H)
## [1] 813
sum(DBFHR$A)
## [1] 593

A warning message appears for this test, apparently it could be because my data set is too small which could lead to errors ( (n.d.)). I decided to play the blind card here and ignore the warning message hoping it won’t have any impact on my results. Now that I “solved” this warning message problem, let’s have a look at the result. In order to interpret it I used a alpha (α) of 5% as it is commonly the case (at least in the psychology domain). The p-value associated is smaller than 5%, which means that I can reject the null hypothesis.Therefore,the number of wins at Home is not the same as the number of wins Away in the Swiss Super League championship. My alternative hypothesis was orientated towards more wins at home, to see if this is the case I summed up the number of games in the column Home and Away. The number of wins at home is bigger than the number of wins Away.

Coming to an end

Findings

As a good report, it’s now time to conclude. Thanks to the different manipulation I did with my initial data I was able to answer my research question. The statistics show that in the Swiss Super League the number of wins at Home is statistically greater than the number of wins Away. This finding fits with my initial thoughts on the topic and the Home Advantage phenomena.

Servette FC

As a proud Servette FC fan I now know that I have a greater probability to see my team win at Home than Away. Or do I ? My statistics were actually made looking at all the teams of the Swiss Super League together, but there could be some outliers. What do I mean by outliers ? Well teams that do not follow the same pattern than the others. I was curious to see if it was the case for Servette and decided to do some quick exploring research. I didn’t adventure again into statistics as I was kind of traumatized from my previous experience, so I won’t be able to say if those results are statistically significant or not.

## 
##  A  D  H 
## 24 27 29

## 
##  A  D  H 
## 21 17 42

Those quickly made ( I actually spend like an 1 hour to be able to chose the color from the pie chart lol) tables and graphs might not be statistics but they still inform about some information. For example we can see that when playing at Home, Servette has won 29 games, and 21 when playing away. It’s hard to say if those results are statisically different without using statistics, but it still give us some information. I think it’s important to raise the fact that even though the number of wins Home and Away are not too different, the number of losses are. Servette lost 24 times at Home and 42 away. Thus, it means that the number of times the match ended up in a draw is greater at home than away. This last bit of information could be interesting to investigate in the future.

Travelling through Time

I could have ended up my work here (especially because it’s getting realllyyyyy late) but I am too curious to do so. I know that the “power” of a team fluctuates with time, so I was curious to see if the number of wins at Home and Away did also. Of course, as it’s something to see team by team, I chose to look at … Servette FC. Here again it’s only some quick data exploration without statistics or anything.

Season 2012/2013 vs 2021/2022

SerHome1213 <-DBF2 %>% filter(DBF2$Season== "2012/2013", DBF2$Home== "Servette")
SerAway1213 <-DBF2 %>% filter(DBF2$Season== "2012/2013", DBF2$Away== "Servette")

SerHome2122 <-DBF2 %>% filter(DBF2$Season== "2021/2022", DBF2$Home== "Servette")
SerAway2122 <-DBF2 %>% filter(DBF2$Season== "2021/2022", DBF2$Away== "Servette")

 table(SerHome1213$Res)
## 
##  A  D  H 
## 10  4  4
 table(SerAway1213$Res)
## 
##  A  D  H 
##  2  4 12
 table(SerHome2122$Res)
## 
## A D H 
## 6 5 7
 table(SerAway2122$Res)
## 
##  A  D  H 
##  5  3 10

The results here tell us that:

In 2012-2013 Servette won only 4 times at Home and 2 Away (it wasn’t a good year lol).

In 2021-2022 Servette won 7 times at Home and 5 Away (it’s getting better).

In both cases the total number of wins at Home is bigger than those Away, but again we need statistics to make sure those results are significant.

I am not sure but I think here it would be interesting to make an anova with Season and Place (Home or Away) as factors.

The End

I could keep looking at this data all night long, but I still have a lot to do and need some sleep. I have found this work really interesting, even though sometimes I just wanted to throw it all away lol. Even though I focused my research on the Swiss Super League, the nice thing about R is that you can re-use it with almost any data. Same goes with my examples with Servette, if you are curious to know how it’s for other teams you just have to change “Servette” by the name of the team you are interested in (that is in this database or changing some code to adapt it to your database).

Even though statistics have been my worst nightmare in this research, I would like to end my report with a nice quotation from my teacher Steven Moran that actually saved me and this work by answering my questions. Here comes the quote:
⁠“The point here is that there’s always different ways of using stats – that’s why although their math is math, their use is basically art.” So in some way I guess I decided to do art by manipulating the way I am going to use my stats.

Behind the scenes

You are bored or don’t know what to do? Well, this little extra part is for you (or you could just skip it).

Here I just copy-paste all the code I wrote that actually work out but that I finally didn’t need. I don’t really know if it’s going to be of some use to someone, but to be honest it made me sad thinking I had to delete it.

Complexicating the simplification

To code the games’ results, I decided to code W (Win) when the HG (Home Goals) was bigger than AG (Away Goals), D (Defeat) when AG was bigger than HG, and t (Tie) when the HG was equal to the AG.

#Wins

W <- DBF2$HG>DBF2$AG
W
table(W)["TRUE"] #813 Win3
table(W)["FALSE"] # 1071 no Wins (Defeats or Ties)

D <- DBF2$HG<DBF2$AG
D
table(D)["TRUE"] #593 Defeats
table(D)["FALSE"] #1291 no Defeats (Wins or Ties)

t <- DBF2$HG==DBF2$AG
t
table(t)["TRUE"] #478 Ties 
table(t)["FALSE"] #1406 no Ties (Wins or Defeats)

#to make sure: 
813+593+478 == 1884

I was kind of proud of myself for finding out a way to code the games depending on the score… all that to discover 3 days before the due date that there was a column called “Res” on my dataset that already did that. Well, at least I learned how to do it, and now I am sure there is no mistake in the column.

How I felt

So if I could have a recommendation for the data scientist community: explore your data, but really, take your time, it will make you save it later.

I wonder why

Even though I start getting use to R Markdown there are some things I still don’t understand. For example, I wanted to center all my Titles, and don’t ask me why but it worked for some of them, but not for others. I am hoping it’s a problem of my computer and not my report, but I think this is too much to hope for. Another thing I didn’t manage to do is create an automatic bibliography. I mean I even had help from my classmate Estelle, who told me what to download and how to do it, but I failed. Even though I have the packages downloaded, no need to hide it, I did it manually, basing it on the APA style. Speaking about downloaded packages, I tried to use a lot of them, but wasn’t always successful, so some might be downloaded but remain unused. By the way, if you have the package “MASS” on you should turn it off when running this work, as one of its functions (select) goes into conflict with one of my other packages (I think it’s dplyr).

References

Chi-Square Test of Independence in R. STHDA. Consulted on December 2022 on http://www.sthda.com/english/wiki/chi-square-test-of-independence-in-r>

Football-Data consulted on December 2022 on https://www.football-data.co.uk

  1. Funny Football Memes 2022 – Kick Off the Season With Humor. Digital mom blog. Consulted on December 2022 on https://www.digitalmomblog.com/football-memes/

  2. SERVETTE FC - FC LUCERNE 1-1 - STADE DE LA PRAILLE. Funny Live. Consulted on December 2022 on https://www.funnylive.com/fr/evenement/servette-fc-fc-lucerne-1-1-stade-de-la-praille-2022-02-20

  3. Statistical test for determining advantage from win ratio in a 4 player game.Stack Exchange. Consulted on December 2022 on <https://stats.stackexchange.com/questions/484631/statistical-test-for-determining-advantage-from-w in-ratio-in-a-4-player-game>

Van den Rul, C. (2019) How to Generate Word Clouds in R. Towards data science. Consulted on December 2022 on https://towardsdatascience.com/create-a-word-cloud-with-r-bde3e7422e8a>

Zach. (2020). How to Create Tables in R (With Examples). Statology. Consulted on December 2022 on https://www.statology.org/create-table-in-r/

n.d. http://mastersportal.com.